Microglia replacement by ER-Hoxb8 conditionally immortalized macrophages provides insight into Aicardi-Goutières syndrome neuropathology

Abstract

Microglia, the brain's resident macrophages, can be reconstituted by surrogate cells - a process termed 'microglia replacement'. To expand the microglia replacement toolkit, we here introduce estrogen-regulated (ER) homeobox B8 (Hoxb8) conditionally immortalized macrophages, a cell model for generation of immune cells from murine bone marrow, as a versatile model for microglia replacement. We find that ER-Hoxb8 macrophages are highly comparable to primary bone marrow-derived macrophages in vitro, and, when transplanted into a microglia-free brain, engraft the parenchyma and differentiate into microglia-like cells. Furthermore, ER-Hoxb8 progenitors are readily transducible by virus and easily stored as stable, genetically manipulated cell lines. As a demonstration of this system's power for studying the effects of disease mutations on microglia in vivo, we created stable, Adar1-mutated ER-Hoxb8 lines using CRISPR-Cas9 to study the intrinsic contribution of macrophages to Aicardi-Goutières syndrome (AGS), an inherited interferonopathy that primarily affects the brain and immune system. We find that Adar1 knockout elicited interferon secretion and impaired macrophage production in vitro, while preventing brain macrophage engraftment in vivo - phenotypes that can be rescued with concurrent mutation of Ifih1 (MDA5) in vitro, but not in vivo. Lastly, we extended these findings by generating ER-Hoxb8 progenitors from mice harboring a patient-specific Adar1 mutation (D1113H). We demonstrated the ability of microglia-specific D1113H mutation to drive interferon production in vivo, suggesting microglia drive AGS neuropathology. In sum, we introduce the ER-Hoxb8 approach to model microglia replacement and use it to clarify macrophage contributions to AGS.

Keywords: immunology; inflammation; leukodystrophies; microglia; microglia replacement; mouse; neuroscience.

Figure 1.. Comparison of ER-Hoxb8 to bone marrow-derived (BMD) macrophages in vitro.

(A) Schematic for creation of BMD and ER-Hoxb8 cells. (B) Brightfield images of BMD and ER-Hoxb8 macrophages plated in the presence of 30 ng/mL mouse CSF1 and differentiated for 7 days (scale bar = 100 um). (C) Dot plot representing CD45/CD11B levels (pre-gated on live, singlet, leukocyte) by flow cytometry. (D) Heatmap showing Log2 CPM of canonical macrophage (top) and non-macrophage (bottom) immune cell genes. (E) Whole-transcriptome comparison between BMD and ER-Hoxb8 macrophages, depicting best fit line and coefficient of determination (one dot = one gene). (F) Volcano plots comparing all genes or those with CPM >1 (Log2FC ≥ 2, FDR <0.05); blue = upregulated in ER-Hoxb8 macrophages, red = upregulated in BMD macrophages. Panel (A) created with https://Biorender.com/w42j197.

Figure 1—figure supplement 1.. Extended validation and comparison of ER-Hoxb8 to bone marrow-derived (BMD) macrophages in vitro, relating to Figure 1.

(A) Gating strategy for Figure 1C. (B) Flow cytometry histograms and median fluorescence intensity (MFI) of CD11B (left) and CD45 (right), relating to Figure 1C. (C) Merged image of ER-Hoxb8 macrophages (green) and synthetic apoptotic cell corpse mimics (10% PS beads, purple) imaged after fixation and staining (orange denotes beads counted as ingested). (D) Quantification of phagocytosis of synthetic apoptotic cell corpse mimics (10% PS beads) as compared to synthetic live cell mimic controls (PC beads); p-value calculated via nested t-test for n=4 replicates per group, with overall distribution (81–100 cells per replicate) shown in gray; ***p<0.001. (E) Heatmap showing Log2 CPM gene expression levels of macrophage genes across groups; progenitor and macrophage groups are as represented in Figure 1A, BM Mono = monocytes isolated from BM, 4div ER-Hoxb8=ER-Hoxb8s collected after 4 days of in vitro differentiation. (F) PCA plot combining bulk RNA sequencing data amongst all groups. (G) Log2 CPM values of top 10 differentially expressed genes by Log2FC (CPM >1, Log2FC ≥ 2, FDR <0.05).

Figure 2.. Engraftment potential of ER-Hoxb8 compared to bone marrow-derived (BMD) macrophages after intracranial transplantation in Csf1r^-/- hosts.

(A) Schematic for in vivo Csf1r^-/- transplant experiments. (B) Rendered tile stitches of Csf1r^-/- brains after intracranial injection of GFP+ bone marrow (left) or ER-Hoxb8 (right) progenitor cells. (C) Percent of total brain area tiled by donor cells; n=5–7 biological replicates per group; each dot = one biological replicate (average area across three matched sagittal sections). (D) Immunostaining of cortical brain region 12–16 days post-intracranial injection (red = IBA1, green = endogenous GFP, blue = DAPI; scale bar = 100 um; inset scale bar = 5 um). (E) Cortical density calculations (cells per mm²) between groups; n=5–7 biological replicates per group; each dot = one biological replicate (average density across three regions of interest across three matched sagittal sections). All p-values calculated via one-way ANOVA with multiple comparisons; ns = not significant or p≥0.05, ****p<0.0001. Panel (A) created with https://BioRender.com/j85e198.

Figure 3.. ER-Hoxb8 macrophages become microglia-like cells (MLCs) after engraftment in the Csf1r^-/- brain.

(A) Histogram of TMEM119 surface staining by flow cytometry (pre-gated on live, singlet, leukocyte, CD45+/CD11B+) for brain-engrafted cells 14 days post-intracerebral transplantation; Mg = WT Microglia, BM = BMD MLCs, Hox = ER-Hoxb8 MLCs. (B) PCA plot comparing in vitro macrophages from Figure 1 with in vivo macrophages; Mg = WT Microglia, BM = BMD MLCs, Hox = ER-Hoxb8 MLCs. (C) Whole-transcriptome comparison between WT microglia, BMD, and ER-Hoxb8 macrophages in vivo, depicting best fit line and the coefficient of determination. (D) Comparison of microglia signature genes (Cronk et al., 2018) depicting best fit line and coefficient of determination. (E) In vitro and in vivo Log2 CPM gene expression of 10 canonical microglia/myeloid genes for bone marrow (left) and ER-Hoxb8s (right) as compared to unmanipulated in vivo microglia.

Figure 3—figure supplement 1.. Extended comparison of ER-Hoxb8 and bone marrow-derived (BMD) macrophages after intracranial transplantation in Csf1r^-/- hosts, relating to Figure 3.

(A) Histogram of TMEM119 surface staining by flow cytometry (pre-gated on live, singlet, leukocyte, CD45+/CD11b+) for seven-day differentiated in vitro BMD and ER-Hoxb8 macrophages. (B) Immunostaining of cortical brain region 12–16 days post-intracranial injection (red = IBA1, green = endogenous GFP, white = P2RY12; blue = DAPI; scale bar = 100 um). (C) Gating strategy for in vivo TMEM119 histogram shown in Figure 3A (pre-gated on live, singlet, leukocyte, GFP) of WT microglia and intracranially transplanted (ICT) cells (BM or ER-Hoxb8s). (D) Unsupervised hierarchical cluster dendrogram, related to Figure 3B (distance method = euclidean; cluster method = complete). (E) Comparison of ‘batch’ by PCA plots for harvest days (left), sorter used (middle), and host mouse sex (right). (F) Venn diagrams comparing the top 200 up- and downregulated genes, as compared to unmanipulated in vivo microglia; asterisk denotes data derived from previously published dataset (Bennett et al., 2018); genes were rank-ordered by absolute expression value, with only the top 1000 being considered for comparison, sorted by Log2FC >2, FDR <0.5, and ordered by Log2FC values. (G) Quantification of GFAP percent area covered; n=3 biological replicates per group; each dot = one biological replicate (average area across three matched sagittal sections); p-values calculated via one-way ANOVA with multiple comparisons; ns = not significant or p≥ 0.05. Panel (F) created with https://Biorender.com/2k8tqf3.

Figure 4.. Adar1 mutation prevents macrophage-lineage cell expansion and causes interferon induction, rescued by JAKi or Ifih1 mutation.

(A) Schematic of ADAR1 locus, depicting exons, alternative start sites for p150 and p110 isoforms, and sgRNA targets. (B) ER-Hoxb8 cell counts over differentiation time course. (C) Immunostaining of in vitro, 8-day differentiated macrophages comparing control (NTC) and Adar1 guide-transduced macrophages (red = CD11B, blue = DAPI; scale bar = 100 um) (D) PCA plot of progenitors and macrophages in vitro (E) Volcano plots showing differentially expressed genes between Adar1 KO and NTC progenitors and macrophages (CPM >1, Log2FC ≥ 2, FDR <0.05). (F) Heatmap showing the Log2FC (Adar1 KO values over NTC values) for relevant interferon-stimulated genes. (G) Heatmap showing Log2 CPM of canonical macrophage (top) and non-macrophage (bottom) immune cell genes. (H) Heatmap showing Log2FC (Adar1 KO over NTC expression) for interferon-stimulated genes in macrophages treated with baricitinib. (I) ER-Hoxb8 cell counts over differentiation time course, comparing the effect of baricitinib on Adar KO and NTC lines (dosages = 0 uM, 0.00064 uM, 0.16 uM, 0.4 uM, and 10 uM); statistics calculated at the 9-day time point. (J) Interferon, cytokine, and chemokine production after treatment with baricitinib via cytokine bead array. (K) ER-Hoxb8 cell counts over differentiation time course, comparing NTC, Adar1 KO, and Adar/Ifih1 double KO (dKO) lines – dotted NTC and Adar1 KO lines are equivalent to those shown in panel (B). (L) Interferon, cytokine, and chemokine production via cytokine bead array, comparing NTC, Adar KO, and Adar/Ifih1 dKO lines. All p-values calculated via one-way (L) or two-way (B, I, J, K) ANOVA with multiple comparisons; ns = not significant or p≥0.05, *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001.

Figure 4—figure supplement 1.. CRISPR-Cas9 editing of ER-Hoxb8 progenitors, relating to Figures 4 and 5.

(A) Schematic for Tlr4 knockout using Cas9^+/- ER-Hoxb8s and sgRNA viral transduction. (B) Post-editing trace decomposition charts (via TIDE Analysis) for Tlr4 KO macrophages. (C) Histogram of TLR4 surface staining by flow cytometry (pre-gated on live, singlet, leukocyte) for 8-day-differentiated in vitro BMD and ER-Hoxb8 macrophages. (D) TLR4 median fluorescence intensity (MFI) normalized to FMO levels using flow cytometry data shown in (C). (E) TNFa production, comparing control, LPS, and R848 treated samples (8-day differentiation, 9.5 h LPS or R848, 100 ng/mL, n=4 replicates per condition); p-values calculated via two-way ANOVA with multiple comparisons; ns = not significant or p≥0.05, ****p<0.0001. Panel (A) created with https://Biorender.com/q68t656.

Figure 4—figure supplement 2.. Evidence for gene knockout (KO), Adar1 sgRNA #2 results, and extended bead array data, relating to Figure 4.

(A) Heatmap of Log2 CPM expression of seven causal AGS genes in BMD and ER-Hoxb8 macrophages. (B) Post-editing trace decomposition charts (via TIDE Analysis) for Adar1 KO cells targeted with two distinct guides. (C) ER-Hoxb8 cell counts over time. (D) Immunostaining of in vitro, 8-day-differentiated macrophages comparing control (NTC) and Adar1 sgRNA #2-transduced macrophages (red = CD11B, blue = DAPI; scale bar = 100 um). (E) Heatmap showing the Log2FC (Adar1 KO values over NTC values) for relevant interferon-stimulated genes for both progenitors and macrophages at baseline. (F) Validation of in vitro ER-Hoxb8 cell counts with vehicle or 10 uM baricitinib treatment using Adar1 sgRNA #2; statistics calculated at the 9-day time point. (G) Interferon, cytokine, and chemokine production for ER-Hoxb8 macrophages grown with or without baricitinib, via cytokine bead array. (H) Trace decomposition charts (via TIDE Analysis) for Adar1/Ifih1 double KO (dKO) cell lines. (I) Interferon, cytokine, and chemokine production for ER-Hoxb8 macrophages with or without subsequent Ifih1 KO. All p-values calculated via one-way (I) or two-way (C, F, G) ANOVA with multiple comparisons; ns = not significant or p≥0.05, *p<0.05, **p<0.01, ***p<0.001, ****;p<0.0001.

Figure 5.. Adar1 mutation prevents ER-Hoxb8 engraftment in the Csf1r^-/- mouse, partially rescued by Ifih1 deletion.

(A) Representative rendering of donor cell engraftment (scale bar = 1000 um) with inset microscopy of GFP+ donor cell engraftment (green = endogenous GFP, blue = DAPI; scale bar = 100 um) for control cells (TLR4 KO and NTC) harvested 7–15 days post-injection (dpi), (B) Adar1 KO (sgRNA #1) cells (harvest details in E), (C) Adar1 KO (sgRNA #2) cells harvested 9-12dpi, and (D) Adar1/Ifih1 double KO (dKO) cells harvested 10–15 dpi (rendered dots in two right brains enlarged 5x for visualization). (E) Percent of total brain area tiled with cells between groups (numbers denote ‘n’ per group); Adar1 KO (#1) cells include pooled data (brains injected with 300k cells/hemisphere, harvested at 10–15 days post-injection (dpi; n=3); brains injected with 300k cells/hemisphere, harvested at 4–8 dpi (n=8); brains injected with 50k cells/hemisphere, harvested at 13 dpi (n=4); and brains injected with 100k cells/hemisphere pre-treated with 0.5 uM Baricitinib, mice treated daily with 1 mg/kg Baricitinib, harvested at 5 dpi (n=1)); asterisk indicates samples where engraftment is present but does not meet criteria for tiled brain area, as exemplified in (D); p-values calculated via one-way ANOVA with multiple comparisons; ns = not significant or p≥ 0.05, ***p<0.001, ****p<0.0001.

Figure 5—figure supplement 1.. Concurrent Adar1, Ifih1 mutation morphology, plus engraftment in the Cx3cr1CreERT; Csf1rfl/fl mouse, relating to Figure 5.

(A) Violin plot of all analyzed cortical ER-Hoxb8 cells (n=20) per animal (n=2), with yellow line denoting the median, for number of primary branches, average length of branches, longest branch length, and soma size; p-value calculated via nested t-test for n=2 animals per group; ns = not significant or p≥ 0.05, **p<0.005. (B) Schematic for in vivo Cx3cr1CreERT; Csf1rfl/fl transplant experiments. (C) Representative rendering of donor cell engraftment (scale bar = 1000 um) with inset microscopy of GFP+ donor cell engraftment (green = endogenous GFP, blue = DAPI; scale bar = 100 um) for control cells (TLR4 KO, NTC) harvested 15 days post-injection, (D) Adar1 KO cells harvested seven days post-injection, (E) Adar1/Ifih1 double KO (dKO) cells harvested 13–15 days post-injection, and (F) Adar1 KO cells harvested 3 days post-injection. Panel (B) created with https://Biorender.com/n75g136.

Figure 6.. Adar1 D1113H mutant ER-Hoxb8 macrophages persistently drive brain ISG expression.

(A) In vitro ER-Hoxb8 cell counts over time (p-values calculated via two-way ANOVA with multiple comparisons). (B) Multiplex bead array data for interferons, cytokines, and chemokines produced via ER-Hoxb8 macrophages (p-values calculated via independent two-sample t-test). (C) Sagittal sections of non-transplanted (tamoxifen [tam] sham control) Cx3cr1CreERT; Csf1rfl/fl brains (left) and Adar1 D1113H mutant brains (right) at age P15 and P28-31; nuclei (blue, DAPI), Isg15 (white via RNA in situ hybridization [ISH]); scale bar = 1000 um; red arrow depicts location of corresponding closeup images below, showing IBA1 (red, protein stain), Cre (green, ISH), Isg15 (purple, ISH), and nuclei (teal, DAPI); scale bar = 20 um; see Figure 6—figure supplement 2A for further corresponding closeup images. (D) Sagittal sections of Cx3cr1CreERT; Csf1rfl/fl brains intracranially transplanted with WT ER-Hoxb8s (left) and Adar1 D1113H ER-Hoxb8s (right) at 13 and 27 days post-injection (dpi); nuclei (blue, DAPI), Isg15 (white, ISH); scale bar = 1000 um; red arrow depicts location of corresponding closeup images below, showing IBA1 (red, protein stain), Cre (green, ISH), Isg15 (purple, ISH), and nuclei (teal, DAPI); scale bar = 20 um; see Figure 6—figure supplement 2A for further corresponding closeup images. All p-values characterized by: ns = not significant or p≥0.05, *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001.

Figure 6—figure supplement 1.. Evidence for Cre expression, relating to Figure 6.

(A) Dual immunostaining/RNA in situ hybridization (ISH) showing IBA1 (red, protein stain), Cre (green, ISH), and nuclei (blue, DAPI); scale bar = 100 um (B) magnified image of a repopulated microglia from the non-transplanted (tam sham control) Cx3cr1CreERT; Csf1rfl/fl brain shown in (A); scale bar = 10 um.

Figure 6—figure supplement 2.. Extended Isg15 RNA in situ hybridization (ISH), relating to Figure 6.

(A) Dual immunostaining/RNA ISH showing IBA1 (red, protein stain), Isg15 (purple, ISH), and nuclei (teal, DAPI); scale bar = 100 um; white arrow depicts location of corresponding close up image; scale bar = 20 um. (B) Representative rendering of donor cell engraftment in the Csf1r^-/- brain (scale bar = 1000 um) with inset microscopy of IBA1+ donor cell engraftment (red = IBA1, blue = DAPI; scale bar = 100 um) for FVB Osb-GFP WT ER-Hoxb8s and Adar1 D1113H mutant ER-Hoxb8s harvested 13–14 days post-injection. (C) Dual immunostaining/RNA ISH for samples shown in (B); tile scale bar = 1000 um; inset = Isg15 (pink, ISH), nuclei (blue, DAPI); scale bar = 100 um.